Impact of Piperacillin-Tazobactam Dosing in Septic Shock Patients Using Real-World Evidence: An Observational, Retrospective Cohort Study

Abstract

Background:

Sepsis and septic shock are associated with significant morbidity and mortality. Rapid initiation of appropriate antibiotic therapy is essential, as inadequate therapy early during septic shock has been shown to increase the risk of mortality. However, despite the importance of appropriate antibiotic initiation, in clinical practice, concerns for renal dysfunction frequently lead to antibiotic dose reduction, with scant evidence on the impact of this practice in septic shock patients.

Objective:

The purpose if this article is to investigate the rate and impact of piperacillin-tazobactam dose adjustment in early phase septic shock patients using real-world electronic health record (EHR) data.

Methods:

A multicenter, observational, retrospective cohort study was conducted of septic shock patients who received at least 48 hours of piperacillin-tazobactam therapy and concomitant receipt of norepinephrine. Subjects were stratified into 2 groups according to their cumulative 48-hour piperacillin-tazobactam dose: low piperacillin-tazobactam dosing (LOW; <27 g) group and normal piperacillin-tazobactam dosing (NORM; ≥27 g) group. To account for potential confounding variables, propensity score matching was used. The primary study outcome was 28-day norepinephrine-free days (NFD).

Results:

In all, 1279 patients met study criteria. After propensity score matching (n = 608), the NORM group had more median NFD (23.9 days [interquartile range, IQR: 0–27] vs 13.6 days [IQR: 0–27], P = 0.021). The NORM group also had lower rates of in-hospital mortality/hospice disposition (25.9% [n = 79] vs 35.5% [n = 108]), P = 0.014). Other secondary outcomes were similar between the treatment groups.

Conclusions and Relevance:

In the propensity score–matched cohort, the NORM group had significantly more 28-day NFD. Piperacillin-tazobactam dose reduction in early phase septic shock is associated with worsened clinical outcomes. Clinicians should be vigilant to avoid piperacillin-tazobactam dose reduction in early phase septic shock.

Sepsis and septic shock (SS) are highly prevalent in critical care units, with mortality rates of greater than 10% and 40%, respectively.¹ Septic shock is also a global problem and is estimated to account for more than 6 million deaths annually.² Rapid identification and management of sepsis are essential, and inappropriate antibiotic therapy during the first 48 hours of treatment has been shown to increase the risk of mortality by nearly 60%.³ Risk factors associated with suboptimal antimicrobial dosing in SS include disease-driven alterations in pharmacokinetic (PK) and pharmacodynamic (PD) parameters such as volume of distribution and clearance. Clinical practice guidelines for managing sepsis and SS recommend broad-spectrum antibiotic coverage with aggressive antibiotic dosing to maximize antibiotic exposure. One of the most used antibiotics in SS is piperacillin-tazobactam due to its efficacy against Pseudomonas aeruginosa, the primary gram-negative pathogen of concern in SS.⁴ While used off-label for sepsis and SS, the Food and Drug Administration (FDA)-approved dosing label for piperacillin-tazobactam recommends a dose reduction in patients with renal impairment.⁵ Since renal impairment and acute kidney injury (AKI) often occur in SS, dose reductions with piperacillin-tazobactam are frequently indicated based on product labeling. While the importance of rapid antibiotic initiation is well-known, currently no quantitative data analyze the incidence and effects of piperacillin-tazobactam dose reduction in early phase SS patients, thus leaving significant gaps in our understanding of how this practice affects clinical outcomes. Therefore, there is a critical need to understand better the relationship between reduced antibiotic doses and clinical outcomes in early phase SS. We conducted a study to investigate the rate and impact of piperacillin-tazobactam dose adjustment in early phase SS patients using real-world electronic health records (EHR) data.

Materials and Methods

Ethics Approval

The Institutional Review Board at the University of Florida (UF) approved the study as exempt and waived the requirement for informed consent (study # IRB201901482). The article was prepared following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement.⁶

Study Design

This study was a retrospective observational cohort study of large-scale EHR data from UF Health–affiliated hospitals. Subjects were identified using data obtained through an electronic query of the UF Integrated Data Repository, a health care data repository providing patient-level electronic health data from UF Health patients. The study used EHR clinical data from admitted hospital inpatients from database inception (ie, January 1, 2012) through June 30, 2019. Study data were collected and managed using REDCap electronic data capture tools hosted at the UF. Potential study patients were identified using the predefined inclusion criteria: International Classification of Diseases (ICD) diagnosis codes specific for SS using validated methods,⁷ bar-coded medication administration (BCMA) of piperacillin-tazobactam for at least 48 hours, and BCMA of norepinephrine. Time-zero was defined as the first BCMA receipt of piperacillin-tazobactam. Antibiotic use and other aspects of sepsis management at the study institution were conducted following contemporaneous international treatment guidelines.^8,9 Exclusion criteria were death or hospice discharge disposition within 24 hours of time-zero or receipt of extended-infusion piperacillin-tazobactam during early phase SS. Early phase SS was defined as the first 48 hours of antibiotic treatment. Study patients were categorized into 1 of 2 cohorts based on cumulative piperacillin-tazobactam received during the early phase of SS: normal piperacillin-tazobactam dosing (NORM) group was defined as patients who received ≥27 g of piperacillin-tazobactam during early phase SS (ie, equivalent to at least 3.375 g every 6 hours), and the low piperacillin-tazobactam dosing (LOW) group were patients who received <27 g during early phase SS.

The primary study outcome was the number of norepinephrine-free days (NFD) at study day 28, defined as any day alive and free of norepinephrine through study day 28. All nonsurvivors were assigned zero NFDs as described by Russell and colleagues.¹⁰ Secondary outcomes assessed include 28-day hospital-free days (HFD), in-hospital mortality/hospice disposition, antipseudomonal escalation (APE) rate, and time to APE. APE was defined as piperacillin-tazobactam re-initiation after 24 hours of initial discontinuation or starting an alternative antipseudomonal beta-lactam antibiotic within the 28-day study period. Alternative antipseudomonal beta-lactams included cefepime and meropenem, which were on the institution’s formulary.

Statistical Analysis

Propensity score matching (PSM) was used to account for baseline confounding variables. Pairs were matched according to age, baseline variables (Charlson Comorbidity Index, Rothman Index [RI], estimated creatinine clearance [CrCl]), and receipt of concomitant therapies (empiric combination antibiotic therapy, hydrocortisone, continuous renal replacement therapy [CRRT]). Empiric combination antipseudomonal antibiotic therapy was defined as concomitant aminoglycoside or fluoroquinolone therapy during early phase SS. Estimated CrCl was assessed using the Cockcroft-Gault method using lab variables obtained at time zero.¹¹ Data were censored for patients alive and discharged from the hospital before study day 28. For consequential missing continuous data, stochastic regression imputation was used. Continuous data, including the primary study outcome, were analyzed using the Wilcoxon-Mann-Whitney test. Categorical data were analyzed using a Pearson χ² test. To perform time-to-event analysis, 28-day Kaplan-Meier curves for in-hospital/hospital disposition, time to norepinephrine discontinuation, and time to APE were compared using a log-rank test. A Cox proportional hazards regression model was used to evaluate the influence of potential confounding factors on 28-day inhospital mortality/hospice disposition. Variables included in the model included age, RI score <40, Charlson Comorbidity Index >5, receipt of CRRT, baseline CrCl <40 mL/min, use of hydrocortisone, and receipt of combination empiric antipseudomonal antibiotic therapy. A subgroup analysis of the primary outcome was performed based on baseline renal function. Based on previous literature,^10,12 to detect a 4.5-day difference in NFD and to achieve 80% power, a sample size of 265 patients per group was needed. A 2-tailed P value of ≤0.05 was considered statistically significant. IBM SPSS Statistics for Windows, version 26 (IBM Corp., Armonk, NY, USA) was used for all statistical analysis.

Results

A total of 3833 patients were evaluated for study inclusion, with 1279 patients meeting study criteria. The most common reason for study exclusion was the receipt of piper-acillin-tazobactam less than 48 hours. Among patients meeting study criteria, 928 (72.6%) patients were stratified to the NORM group and 351 (27.4%) belonged to the LOW group. The missing data rate for variables used in the primary and secondary outcomes was less than 5%. After PSM, a total of 608 patients (304 pairs) were included. As expected, patients in the LOW group had a lower 48-hour cumulative dose of piperacillin-tazobactam compared with the NORM group (LOW: 20.3 g, interquartile range [IQR]: 16.9–23.6; NORM: 30.4 g, IQR: 30.4–33.8; P < 0.001). Among patients in the NORM group, 235 (77.3%) of 304 had a 48-hour cumulative dose of piperacillin-tazobactam between 27 and 35.9 g and 69 (22.7%) of 304 had greater than or equal to 36 g (equivalent to 4.5 g every 6 hours).

Baseline characteristics including RI, Charlson Comorbidity Index, hydrocortisone use, receipt of empiric combination antipseudomonal antibiotics, and receipt of CRRT were similar between the treatment groups. Among PSM patients who received empiric combination antipseudomonal antibiotics, the most commonly used antibiotic classes were fluoroquinolones (LOW: 25/304; NORM: 25/304) and aminoglycosides (LOW: 8/304; NORM: 5/304). Baseline characteristics for all patients and PSM patients are included in Table 1.

Table 1.

Baseline Study Characteristics.

Variable	All patients (n = 1279)a			Propensity score matched (n = 608)a
	LOW (n = 351)	NORM (n = 928)	P value	LOW (n = 304)	NORM (n = 304)	P value
Age (years)	65 (55–75)	61 (50–70)	<0.001	65 (55–76)	64 (56–70)	0.256
Male sex	199 (56.7)	572 (61.6)	0.108	171 (56.3)	178 (58.6)	0.566
Admission source
Hospital transfer	163 (46.4)	423 (45.6)	0.802	147 (48.3)	147 (48.3)	1.000
Non-health care facility (eg, home, accident scene)	127 (36.2)	354 (38.1)	0.561	108 (35.5)	105 (34.5)	0.865
Scheduled surgical procedure	13 (3.7)	38 (4.1)	0.873	13 (4.3)	12 (3.9)	1.000
Skilled nursing/intermediate care facility	13 (3.7)	34 (3.7)	1.000	11 (3.6)	14 (4.6)	0.684
Other	35 (9.9)	79 (8.5)	0.442	25 (8.2)	26 (8.6)	1.000
Baseline Rothman Index score ≤40	127 (36.2)	324 (34.5)	0.694	114 (37.5)	105 (34.5)	0.499
Charlson Comorbidity Index score ≥5	168 (47.9)	247 (26.6)	<0.001	154 (50.7)	132 (43.4)	0.088
Body mass index (kg/m2)
< 18.5	26 (7.4)	76 (8.2)	0.723	23 (7.6)	29 (9.5)	0.469
18.5–24.9	85 (24.2)	258 (27.8)	0.204	73 (24.0)	77 (25.3)	0.778
25–29.9	97 (27.7)	227 (24.5)	0.250	81 (26.6)	64 (21.0)	0.128
30–39.9	102 (29.1)	260 (28)	0.729	89 (29.3)	93 (30.6)	0.791
≥40	41 (11.7)	107 (11.5)	0.922	38 (12.5)	41 (113.5)	0.810
Baseline creatinine clearance (mL/min)
<20	73 (20.8)	58 (6.3)	<0.001	105 (34.5)	73 (24.0)	0.006
20–40	126 (35.9)	254 (27.4)	0.003	110 (36.2)	121 (39.8)	0.403
41–60	98 (27.9)	301 (32.4)	0.137	41 (13.5)	72 (23.7)	0.002
61–90	30 (8.5)	151 (16.3)	<0.001	27 (8.8)	9 (3.0)	0.003
91–130	14 (4.0)	99 (10.6)	<0.001	12 (3.9)	19 (6.3)	0.269
> 130	10 (2.8)	65 (7.0)	0.005	9 (3.0)	10 (3.3)	1.000
Time on norepinephrine, days	1.71 (1.0–4.8)	1.00 (1.0–3.5)	0.045	1.71 (1.0–5.0)	1.57 (1.0–6.5)	0.281
Receipt of hydrocortisone	113 (32.2)	245 (26.4)	0.043	100 (32.9)	98 (32.2)	0.931
Receipt of combination antipseudomonal antibiotics	54 (15.4)	100 (10.8)	0.027	51 (16.8)	44 (16.0)	0.503
Cumulative 48-hour piperacillin-tazobactam dose (g)	20.3 (16.9–23.6)	30.4 (30.4–36)	<0.001	20.3 (16.9–23.6)	30.4 (30.4–33.8)	<0.001
Receipt of continuous renal replacement therapy during	103 (29.3)	89 (9.6)	<0.001	82 (26.9)	72 (23.7)	0.401

Italicized P value indicates P value <0.05.

Abbreviations: LOW, low dose of piperacillin-tazobactam; NORM, normal dose of piperacillin-tazobactam.

^aData presented as median and interquartile range or as count and percentage.

Among all patients, NFDs were higher in the NORM group as compared to the LOW group. When evaluating the propensity score–matched cohort, the NORM group had statistically more median NFD, as compared to the LOW group (23.9 days [IQR: 0–27] vs 13.6 days [IQR: 0–27], P = 0.021). In addition, the time to norepinephrine discontinuation was significantly decreased in the NORM group (Supplemental Figure 1). In the subgroup analysis based on baseline renal function, among the propensity score–matched cohort, NORM patients had numerically more median NFDs across all renal function groupings. However, no individual renal function group met the threshold for statistical significance. Table 2 describes the primary outcome results.

Table 2.

Primary Outcome by Baseline Renal Function.

Renal function group (mL/min)	All patients (n = 1279)a			Propensity score-matched patients (n = 608)a
	LOW	NORM		LOW	NORM
	NFD (IQR), n	NFD (IQR), n	P value	NFD (IQR), n	NFD (IQR), n	P value
All patients	14.4 (0–27), n = 351	26.3 (0–27), n = 928	<0.001	13.6 (0–27), n = 304	23.9 (0–27), n = 304	0.021
<20	2.9 (0–26.6), n = 120	26.9 (0–27), n = 144	0.018	22.8 (0–26.9), n = 105	26.9 (0–27), n = 73	0.140
20–40	22.9 (0–27), n = 132	26.5 (0–27), n = 301	0.069	19.4 (0–27), n = 110	22.1 (0–27), n = 121	0.777
41–60	8.4 (0–26.9), n = 47	25.6 (0–27), n = 167	0.006	6.9 (0–26.8), n = 41	22.8 (0–27), n = 72	0.358
61–90	0 (0–25.8), n = 28	26.1 (0–27), n = 152	0.016	1.9 (0–26), n = 27	20.7 (0–26.3), n = 10	0.933
91–130	7.1 (0–25.2), n = 14	26.4 (0–27), n = 99	0.050	7.1 (0–25.2), n = 12	25.9 (0–27), n = 18	0.818
>130	25.8 (3.2–27), n = 10	26.6 (0–27), n = 65	0.768	25.8 (3.2–27), n = 9	26.1 (5.0–26.9), n = 1C	I 1.000

Italicized P value indicates P value <0.05.

Abbreviations: LOW, low dose of piperacillin-tazobactam; NFD, norepinephrine-free days; NORM, normal dose of piperacillin-tazobactam.

^aData presented as median and interquartile range (IQR).

Patients in the NORM group also had lower rates of inhospital mortality/hospice disposition than the LOW group (25.9% [n = 79] vs 35.5% [n = 108]), P = 0.014). Kaplan-Meier curves for estimated survival showed significant differences in in-hospital mortality/hospice disposition through study day 28, P = 0.007 (Figure 1). Using a Cox proportional hazards regression model that included age, receipt of empiric combination antibiotics, receipt of hydro-cortisone, baseline CrCl, baseline Charlson Comorbidity Index, baseline RI score, and receipt of CRRT therapy, the difference in in-hospital mortality/hospice disposition remained statistically significant (Supplemental Table 1). Other secondary outcomes such as APE rate and time to antibiotic escalation (Figure 2) were similar between the treatment groups. Table 3 describes the secondary outcome results.

Figure 1.

Kaplan-Meier curve for 28-day survival/hospice disposition.

Abbreviations: LOW, the low piperacillin-tazobactam dosing group; NORM, normal dose piperacillin-tazobactam.

Figure 2.

Kaplan-Meier curve for time to antibiotic escalation.

Abbreviations: LOW, the low piperacillin-tazobactam dosing group; NORM, normal dose piperacillin-tazobactam.

Table 3.

Secondary Study Outcomes.

Outcome	All patients (n = I279)a			Propensity score-matched patients (n = 608)a
	LOW (n = 351)	NORM (n = 928)	P value	LOW (n = 304)	NORM (n = 304)	P value
Hospital-free days	0 (0–12)	0 (0–14)	0.007	0 (0–11)	0 (0–11)	0.455
In-hospital mortality/hospice disposition	I24 (35.3)	2I9 (23.6)	<0.001	I08 (35.5)	79 (25.9)	0.014
Antipseudomonal escalation rate	7I (20.2)	196 (21.1)	0.794	59 (19.4)	63 (20.7)	0.761

Italicized P value indicates P value <0.05.

Abbreviations: LOW, low dose of piperacillin-tazobactam; NORM, normal dose of piperacillin-tazobactam.

^aData presented as median and interquartile range or as count and percentage.

Discussion

In patients with SS, early administration of broad-spectrum antibiotics is key to achieving positive clinical outcomes. Antibiotics with activity against Pseudomonas aeruginosa are widely recommended for patients with SS, with piper-acillin-tazobactam serving as a frequently used antibiotic due to its high efficacy and tolerability, wide availability, long history of use, and relatively low daily cost, as compared to other antibiotics. While piperacillin-taxobactam can be administered as an extended or continuous infusion, traditionally, piperacillin-tazobactam is administered as an intermittent infusion every 6 hours, with daily doses ranging from 13.5 to 18 g. In a population PK study of 15 patients in early phase SS, Öbrink-Hansen and colleagues¹³ noted that when using traditional piperacillin-tazobactam dosing, patients with preserved or augmented renal function were at higher risk for not reaching PK/PD treatment targets. Our study expands on this work and showed that regardless of the renal function, patients who received lower doses of piperacillin-tazobactam within the initial 48 hours of treatment had less NFD than patients who received higher piperacillin-tazobactam doses. This finding suggests that the use of standard piperacillin-tazobactam dosing yields improved clinical results as compared to reduced-dose piperacillin-tazobactam in early phase SS. To our knowledge, this is the first study to analyze the clinical impact of piperacillin-tazobactam dosing in patients with SS utilizing real-world data, and the results illustrate the importance of aggressive piperacillin-tazobactam dosing early in patients with SS. NFDs were significantly higher among the propensity score–matched cohort; however, in a subgroup analysis based on baseline renal function, no individual renal function category reached statistical significance. This interesting subgroup observation is at risk for a type II error due to the relatively small number of patients in each renal function category. We chose to use NFD as the primary study outcome as opposed to 28-day mortality. NFD has been described as an alternative outcome to mortality in SS trials and has been described as a surrogate for clinical improvement during SS.¹⁰ Furthermore, vasopressor therapy, as a proxy for short-term cardiovascular dys-function, is associated with long-term mortality in SS.¹⁴

Potential explanations for our observed results are varied. There is substantial evidence that critically ill patients, including those with SS, have physiological changes that can significantly affect the PKs of many drugs used in this population. Among these PK changes, increases in the volume of distribution have been described, with shifts in body fluids and changes to plasma protein binding most often implicated. Hydrophilic drugs such as piperacillin-tazobactam are particularly vulnerable to being impacted by changes to the volume of distribution. When the volume of distribution of a drug is increased, an increase in dose is needed to achieve similar serum concentrations compared to when the volume of distribution is normal. Despite this conceptual framework, dosing of medications, including hydrophilic medications, is frequently decreased in the critical care environment. The primary concern driving these dose adjustments is the concern of drug accumulation in the setting of decreased renal function. Pharmacy-managed renal dosing of medications, including antibiotics, is a standard service performed at many hospitals and is generally considered a best practice.¹⁵ At our institution, renal dosing guidelines recommend piperacillin-tazobactam dose adjustment consistent with the FDA-approved drug labeling, recommending an initial dose reduction at a CrCl <40 mL/min and further dose reduction at <20 mL/min, respectively.

Interestingly, nearly half of all patients in the LOW group had better baseline renal function than the FDA-approved threshold for renal dose reduction (n = 142/351; 40.5%). In addition to the potential for unnecessary dose reduction, this may also indicate that factors beyond renal function may influence antibiotic dose reduction in early phase SS. Delays in follow-up antibiotic doses^16–18 or adequate intravenous access¹⁹ can also negatively affect the timely administration of antibiotics, thus contributing to antibiotic dose reduction in early phase SS. Our study was not designed to evaluate these potential factors, but they warrant consideration when addressing causes of reduced antibiotic dosing in early phase SS. However, regardless of the reasons, our study demonstrates the implications of piperacillin-tazobactam dose reduction in patients with SS, particularly early in the disease course.

In addition, our study is the first to our knowledge to use large-scale real-world evidence to quantify the extent to which antibiotic dose reductions occur during the early phases of SS. In this study, nearly a third of all patients received less than the equivalent of the standard FDA-approved dose of piperacillin-tazobactam. In a single-center study of 189 hospitalized patients in Saudi Arabia, underdosing of antibiotics compared with tertiary dosing references was noted in 40% of patients.²⁰ Our results are consistent with this finding and illustrate that underdosing of piperacillin-tazobactam occurs in a significant portion of patients with SS, despite clinical practice guidelines which recommend aggressive antibiotic dosing in this patient population. Beta-lactams such as piperacillin-tazobactam are eliminated predominantly by the kidneys and have wide therapeutic indices with relatively few safety concerns compared to other antibiotic classes. The potential benefits of a deferred antibiotic dose reduction strategy in patients with AKI have been described.²¹ While further research is needed to confirm if our observed study results apply to other beta-lactams used in sepsis and SS, it would appear to be pragmatic to avoid beta-lactam dose reduction in patients with early phase SS.

One particular strength of this study is the use of PSM to account for potential confounding factors that may influence NFDs. Before PSM, imbalances were noted in patient age, baseline Charlson Comorbidity Index, baseline renal function, receipt of hydrocortisone, receipt of combination empiric antibiotic therapy, and receipt of CRRT therapy. After PSM, these imbalances were largely eliminated. These variables were included as they have been associated with worsened sepsis outcomes. In addition to being well-matched based on these characteristics, the propensity score–matched groups were also similar in admission source and body mass index. Another strength of the study was the use of the RI, an artificial intelligence/machine-learning generated severity of illness tool integrated with the EHR. The RI uses 26 variables, including vital signs, laboratory data, cardiac rhythms, and nursing assessments. RI scores range from −91 to 100, with lower scores indicating more severe illness. An RI score of less than 40 identifies patients at high risk for poor outcomes and is associated with overall mortality ranging from 17% to 25%.^22–24 While used in this study to retrospectively assess the severity of illness among study patients, the use of machine learning and artificial intelligence screening tools such as the RI for real-time risk stratification are promising.

Our study is not without limitations. Due to its retrospective observational nature, the potential for selection and information bias cannot entirely be excluded. However, our study takes a pragmatic real-world approach to assessing this research question, as it would be unethical to randomize patients with early SS to suboptimal antibiotic doses. In addition, undercoding of sepsis using administrative data has also been reported.²⁵ However, we attempted to minimize selection bias by using a validated ICD-10 administrative coding data method to identify sepsis and SS patients with high sensitivity and specificity.⁷ In addition, the observed 28-day mortality rate in both treatment groups is consistent with previously reported SS mortality,²⁶ suggesting that the risk of selection bias is low.

Our study also includes real-world data collected from a single health system which could limit external validity. However, given that clinical practice guidelines and regulatory requirements typically guide sepsis management, we do not expect significant variation at similar institutions compared to our health system.

In addition, while we have attempted to account for multiple confounders, including age, severity of illness, chronic comorbidities, baseline renal function, concomitant use of combination antibiotics and corticosteroids, and receipt of CRRT, the impact of unmeasured confounding variables cannot be excluded. Lack of access to microbio-logical data, lack of identification of the definitive source of infection, and no assessment of the adequacy of fluid resuscitation and other sepsis treatments used are notable. However, in early phase SS, antibiotics are primarily empiric and are infrequently guided by definitive culture results. In addition, while the source of infection is essential, current sepsis treatment guidelines do not differentiate antibiotic dosing based on the presumed source of infection, instead recommending rapid antibiotic administration for most patients. In addition, while fluid resuscitation is vital to improving outcomes of patients with SS, emerging evidence suggests that appropriate and timely antibiotic initiation is even more critical.²⁷ Furthermore, our study did not evaluate trends in norepinephrine dosing before piperacillin-tazobactam initiation. However, given the recognition of the importance of early antibiotics in the overall treatment of sepsis, it is unlikely that patients were on vasopressor therapy for a substantial time without receiving broad-spectrum antibiotics.

Even though continuous or extended infusion piperacillin-tazobactam dosing has been associated with a favorable PK/PD profile,^28,29 our study included only intermittent infusion administration, which was the primary method used in early phase SS at our health system. However, our findings may have applicability with other piperacillintazobactam administration methods. Li and colleagues compared the population PK/PD of piperacillin-tazobactam among hospitalized patients with complicated intra-abdominal infection³⁰ and found no PK/PD differences among patients receiving intermittent or continuous infusion piper-acillin-tazobactam. This implies that our findings of the importance of adequate piperacillin-tazobactam dosing may also apply to other modes of administration, such as continuous infusion. However, the applicability of our findings to continuous infusion administration is hypothesis-generating, and further research would be needed to confirm these results in that treatment modality or in other modalities such as extended infusion.³¹ In addition, our study did not evaluate the use of therapeutic drug monitoring (TDM)-guided piperacillin-tazobactam therapy, limiting our ability to extrapolate to this treatment method. TDM-guided dosing has been advocated to improve the achievement of PK/PD targets, particularly in critically ill patients. However, despite its promise, TDM-guided dosing of piperacillintazobactam has yet to become the standard of care across US institutions.³²

Finally, the presence of piperacillin-tazobactam-induced adverse effects, including drug-induced seizures, was not assessed. However, the overall incidence of piperacillin-tazobactam-induced seizures is a rare phenomenon.³³ Despite these limitations, our study has applicability to institutions that use piperacillin-tazobactam to treat SS and represents a real-world evaluation of antibiotic use in this population.

Conclusion and Relevance

This study found that reduced piperacillin-tazobactam doses in early phase SS are frequently used, with nearly one-third of patients receiving less than the standard FDA-approved dose. Upon PSM, patients who received lower doses of piperacillin-tazobactam were noted to have a significantly higher risk for worsened clinical outcomes than patients who received standard dosing. Clinicians should exercise caution when reducing piperacillin-tazobactam doses, particularly in early phase SS. Based on our study results, when initiating piperacillin-tazobactam in patients with SS, clinicians should consider utilizing aggressive dosing strategies for at least the first 48 hours of therapy and minimize the potential for dose reduction, regardless of the patient baseline renal function.